Automatic recognition of frame quality degradation for inspection of surveillance camera

When surveillance camera is broken down, it will degrade frame quality directly. Sometimes, quality degradation happens occasionally, it is difficult for people being aware it immediately. With the aim to automatically inspect surveillance camera, we propose an automatic method to recognize frame quality degradation. Seven features are extracted based on four kinds of measures, i.e. mean of structure similarity, variation of intensity difference, minimum of block correlation, and average color. Those measures have different reactions to different degradations. Subsequently, linear discriminant analysis (LDA) applied to the extracted features is able to train classifiers. Six classes of degradations are recognized in this work, including signal missing, color missing, local alternation, global alteration, periodic intensity change, and normal status. After implementing degradation recognition, we determine whether surveillance camera works normally or not. The experiment results demonstrate that the proposed method is capable of recognizing degradation as well as inspecting surveillance camera.